arX

iv:1

307.

3313

v1 [

astr

o-ph

.CO

] 1

2 Ju

l 201

3

Long Term Spectral Evolution of Tidal Disruption Candidates Selected by

Strong Coronal Lines

Chen-Wei Yang1, Ting-Gui Wang1, Gary Ferland2, Weimin Yuan3, Hong-Yan Zhou1,4, Peng

Jiang1

ABSTRACT

We present results of follow-up optical spectroscopic observations of seven rare,

extreme coronal line emitting galaxies reported by Wang et al. (2012) with Multi-Mirror

Telescope (MMT). Large variations in coronal lines are found in four objects, making

them strong candidates of tidal disruption events (TDE). For the four TDE candidates,

all the coronal lines with ionization status higher than [Fe VII] disappear within 5-9

years. The [Fe VII] faded by a factor of about five in one object (J0952+2143) within

4 years, whereas emerged in other two without them previously. A strong increment in

the [O III] flux is observed, shifting the line ratios towards the loci of active galactic

nucleus on the BPT diagrams. Surprisingly, we detect a non-canonical [O III]λ5007/[O

III]λ4959≃2 in two objects, indicating a large column density of O2+ and thus probably

optical thick gas. This also requires a very large ionization parameter and relatively

soft ionizing spectral energy distribution (e.g. blackbody with T < 5 × 104 K). Our

observations can be explained as echoing of a strong ultraviolet to soft X-ray flare caused

by tidal disruption events, on molecular clouds in the inner parsecs of the galactic nuclei.

Re-analyzing the SDSS spectra reveals double-peaked or strongly blue-shouldered broad

lines in three of the objects, which disappeared in the MMT spectra in two objects, and

faded by a factor of ten in 8 years in the remaining object with a decrease in both the line

width and centroid offset. We interpret these broad lines as arising from decelerating

biconical outflows. Our results demonstrate that the signatures of echoing can persist

for as long as ten years, and can be used to probe the gas environment in the quiescent

galactic nuclei.

Subject headings: black hole physics – galaxies: nuclei – line: formation –

1Key Laboratory for Research in Galaxies and Cosmology, The University of Sciences and Technology of China,

Chinese Academy of Sciences, Hefei, Anhui 230026, China (twang@ustc.edu.cn)

2Department of Physics, University of Kentucky, Lexington, KY 40506, USA

3National Astronomical Observatory, Chinese Academy of Sciences,20A Datun Road, Beijing, China

4Polar Research Institute of China, 451 Jinqiao Road, Pudong, Shanghai 200136, China

– 2 –

1. Introduction

Stellar and gaseous kinematics suggests that most galaxies harbor central supermassive black

holes (SMBH) and that the SMBH masses are well correlated with the stellar masses or velocity

dispersions of the galactic bulges (e.g. Kormendy & Richstone 1995; Magorrian et al. 1998; Ho 1998;

Merritt & Ferrarese 2000; Gebhardt et al. 2000). When a star in the galactic nucleus accidentally

passes within the tidal disruption radius of the massive black hole (rp < RT ≃ R∗(MBH/M∗)1/3,

Hills 1975), it is torn apart by the tidal force of the black hole. About half of the stellar debris is

ejected and the rest falls back towards the black hole, causing a bright flare lasting for a few months

to years (Rees 1988; Ayal et al. 2000). For a black hole of mass between 107 M⊙ and 108 M⊙ and a

solar-type star, the peak luminosity will be sub-Eddington (LEdd = 1.3× 1038(MBH/M⊙)erg s−1).

For less massive black hole (MBH < 107 M⊙), the rate of fall back initially exceeds the Eddington

accretion rate, likely launching strong outflows. The spectra of the super-Eddington accretion tori

and outflows are still uncertain, but when the fallback rate is below the Eddington rate, the gas is

accreted onto the black hole via a thin disk and its radiation peaks in the UV to soft X-ray bands.

Besides UV and soft X-ray continuum flares, variable broad and narrow emission lines may

be also seen in the spectra. Part of the UV and X-ray light may be reprocessed by the outflows

or unbounded debris, giving rise to the broad emission lines (Bogdanovic et al. 2004; Strubbe

& Quataert 2009; Wang et al. 2011, hereafter W11; Gezari et al. 2012). The same UV and

X-ray sources may also ionize the cold circum-nuclear interstellar medium (ISM), producing high

ionization narrow lines (Ulmer 1999; Komossa et al. 2008; Wang et al. 2011, 2012). Broad lines

are expected to follow the continuum variations closely because they are produced in a region very

close to the continuum source. Narrow lines should also be variable, but on a longer time scale due

to the response of ISM material further away to the continuum flare (light echoing). In particular,

the high-ionization narrow lines which vary on time scales of years are unique features for the

light echo of tidal disruption event (TDE), that distinguishes it from active galactic nuclei (AGNs),

which possess narrow line regions (NLRs) of sizes from 102 to 103 pc. Because gas at large scale

will have little response to the continuum flare on time scales of months to years, we may expect

to observe only the spectrum from the very inner region of the NLR, which is likely dominated by

high ionization lines, while low ionization lines from normal NLR is weak.

Wang et al. (2012, hereafter W12) carried out a systematic search for extreme coronal line

emitters (ECLEs) in the spectroscopic samples of galaxies and low redshift quasars from the Sloan

Digital Sky Survey (SDSS) Data Release 7 (DR7), following the initial results of detecting variable

narrow lines in two ECLEs (Komossa et al. 2008, 2009; W11). They identified a sample of seven

galaxies that show extremely strong coronal lines from [Fe X] up to [Fe XIV]. Traditional narrow-

line diagnostics suggested they are non-active, and half of the sample show broad emission lines of

complex profiles. Also the sample is split into low and high ionization subclasses according to the

detection or non-detection of [Fe VII]. W12 showed that these lines are formed in a photo-ionized

gas, and argued that ECLEs are most likely the light echoes of a tidal disruption flare based on

energetics considerations. Most these galaxies are intermediate luminosity disk galaxies with small

– 3 –

bulges. If all these sources turn out to be TDE, then the rate of such TDE is expected around a few

times 10−5 per galaxy per year for galaxies in the absolute magnitudes around −21.3 < Mi < −18.9.

However, there are some further questions left from previous studies. First, optical follow-ups

were carried out for only two of the seven objects. It remains to be verified whether the rest of the

sample also show the same variation trends. Second, the long term evolution of the emission lines

is not known. On what time scale do coronal lines disappear? Are there any other special features

in the later stage of evolution that can be used to identify such events on long time scales? With

these questions, we initiated spectroscopic follow-up observations with the Multi-Mirror telescope

(MMT) to characterize the spectral signatures at late phases of evolution. In this paper we will

present an analysis of the follow-up MMT spectroscopic observations. The paper is organized as

follows. We present the observations and data reduction in §2 and results on the spectral evolution

in §3. We discuss the implication of these results in §4. Throughout the paper, we will adopt a

Λ-CDM cosmology with H0 = 71km s−1 Mpc−1, ΩM = 0.28 and ΩΛ = 0.72.

2. Observations and Data Reduction

We observed all seven targets with the Blue Channel Spectrographs mounted on the Multi-

Mirror Telescope (MMT) on December 26, 2011. The log of the observations is given in Table 1.

During the observation, the typical seeing was 0.′′8 and we adopt a long slit with 1′′ width. The slit

is centered on the galactic nuclei and orientated at the parallactic angle. We took exposures of all

seven targets for 900s with 500 gpm grating, centering at 6000A. This setting results in a spectral

resolution R = λ/∆λ = 1430 and a wavelength coverage from 4430A to 7560A. Unfortunately, the

slit was not properly placed on the center of the galaxy for SDSS J1241+4426, thus we will ignore

the red part of spectrum for this object. We also obtained the blue portion of spectra for five

objects, including SDSS J0748+4712, SDSS J0938+1353, SDSS J0952+2143, SDSS J1055+5637,

and SDSS J1241+4426 using 800 gpm grating, centered at 4100A, and an exposure time of 900s for

each. The latter gives a spectral resolution of 1730 and a wavelength coverage from 3100A to 5100A.

Three and one KNPO standard stars were observed using the same settings as 500 gmp and 800

gmp during the observation for flux calibration, and 7 and 5 He-Ne-Ar lamp spectra with g500 and

g800 grism sets were taken for the wavelength calibration. The raw two-dimension data reduction

and spectral extraction were accomplished using standard routines in IRAF. To extract the nuclear

spectra, we used the APALL task1 and choosed an aperture of 3′′. We also extract the spectra

with an aperture of 1′′, all high ionization lines including coronal lines, [O III] and He IIλ4686

are almost the same, so must come from the center 1′′ region. We carried out the wavelength and

flux calibration using the spectra of a He-Ne-Ar lamp and those of the standard stars, respectively.

When applying different standard stars for flux calibration taken with 500 gpm, the spectra of the

same object were consistent with each other within 5%. Thus, we take the medians as our final

1APALL is a multi-step task which defines and extracts the data from 2D CCD image in IRAF.

– 4 –

results. After the flux calibration, the blue and red spectra of the same object taken with the

two gratings are consistent with each other within 8% in the overlapping region, indicating small

uncertainty in the flux calibration. Then we rescale the blue part spectrum to match the red one,

and then combine them. We estimate the spectral resolution by measuring the width of emission

lines in the spectrum of He-Ne-Ar lamp. For 500 gpm grating and 800 gpm grating, their typical

values are 90 km s−1 and 69 km s−1.

3. Results

3.1. Comparison between SDSS and MMT spectra

The MMT spectra are shown in Figure 1. For comparison, we also overlay the SDSS spectra

taken 5-9 years ago. Note that the SDSS spectra were taken using a larger aperture, a circle of

3.′′0 in diameter, than ours, in 1.′′0 × 3.′′0 region, so they contain more light from the host galaxy

than our MMT spectra. As such the low-ionization narrow lines from the extended star-formation

regions may be significantly smaller in the MMT spectra, along with the starlight; while we do not

expect that a significant coronal lines come from the extended region. Therefore, coronal lines will

not be significantly affected by the aperture effect.

An initial check of the Figure 1 suggests two different variability trends for coronal lines.

In four of seven objects, the high-ionization coronal lines [Fe X]λ6376-[Fe XIV]λ5304 completely

disappeared in the MMT spectra, while for other three objects J0938+1353, J1055+5637 and

J1241+4426, the MMT spectra are similar to the SDSS ones with strong high-ionization coronal

lines. For J1241+4426, although we failed to obtain the red portion spectrum of the galactic center,

we find that [Fe VII]λ3759 and [Ne III]λ3896 lines are prominent and do not show significant

variation in the blue portion of the spectrum. Therefore, the latter three objects seem to be AGNs.

The broad Balmer lines, Fe II and the non-stellar continuum remain very prominent in J1055+5637,

thus it is a type 1 Seyfert galaxy. Note that J1055+5637 is the only object in W11 with narrow-

line ratios located in the AGN regime in all three BPT diagrams. J0938+1353 shows fairly broad

coronal and high ionization lines, superposing on strong, much narrower, low ionization lines, which

come clearly from star-formation regions. We suspect that coronal lines and high ionization lines

are from an obscured AGN. Thus it is likely a composite of a Seyfert 2 nucleus with star-formation

regions but its true nature remains mysterious as coronal lines are not usually seen in composite

type galaxies. Alternatively, the coronal lines may be related to tidal disruption of a giant star by

a massive black hole, which produces a much longer lasting flare than of tidal disruption of a main

sequence star. The latter scenario can be tested by future spectroscopic follow-up observations.

We will focus only on the four variable objects in the late analysis.

– 5 –

3.2. Continuum Modeling and Variability

In this subsection we will model the continuum and examine the contribution of potential non-

stellar continua in the SDSS spectrum. W12 found that four of the seven objects show flux variations

between SDSS photometric and spectral observations, including the AGN J1055+5637. For two

objects without [Fe VII] lines, they were brighter during the spectroscopic observations than during

the photometric observation, indicating that there is a non-stellar continuum in J0748+4712 and

J1350+2916 spectra, while it is difficult to assess the case in J0952+2143. If TDE is responsible

for the continuum variability, any non-stellar continuum emission would be very weak after 5-9

years even if it existed during the SDSS spectroscopic observations. Indeed, in all MMT spectra,

the continua are dominated by starlight. Therefore, the continua in the MMT spectra can be used

as first- order templates for the stellar light in the SDSS by ignoring the radial gradients of stellar

populations.

In order to get the starlight spectrum beneath the emission lines and also beyond the MMT

spectral coverage, we fit the MMT spectra with a combination of 6 independent components (ICs)

templates derived from the stellar populations (BC03) with Ensemble Learning Independent Com-

ponent Analysis. Previous tests show that such extrapolation is reliable (see Lu et al. 2006 for

detail). In practice, the continuum fit to a narrow wavelength coverage (4430A to 7560A) of the

MMT spectrum of J1350+2916 can match its SDSS spectrum (wavelength coverage from 3700A to

8500A) well. During the fitting we mask all prominent emission lines, and convolve the templates

with a Gaussian kernel and shift in redshift to match the width and the centroid of the absorption

lines. The fits with 6 ICs are shown also in Figure 1.

Keeping in mind that different amounts of the starlight lie within the apertures, we rescale

the starlight models of J0748+4712, J0952+2143 and J1350+2916 derived from the MMT spectra

to match their SDSS ones by minimizing the residuals in the stellar absorption lines. This yields

the fit in Figure 2. Overall, the stellar absorption lines in the SDSS spectra are fitted very well,

suggesting that this approach is feasible. A non-stellar continuum is required to be present in

only SDSS J0748+4712. This provides an independent confirmation to the continuum variability

analysis, further it also gives an estimate of the non-stellar continuum spectrum. Note this is much

less model dependent than the spectral decomposition method using a combination of starlight plus

a power-law or black-body fit in W11. We will describe the individual object below.

For J0748+4712, its SDSS spectrum shows strong non-stellar emission and three broad bumps

around 4050, 4600 and 6560A (see W11). The scaling factor of the star-light in the SDSS spectrum

is 1.35. After subtracting the stellar light spectrum, the differential spectrum is a smooth continuum

plus several broad bumps and narrow emission lines (Figure 2). One noticeable difference from W11

is that there is no broad feature around 4050A, suggesting it was due to imperfect subtraction of

stellar light in W11. The continuum is blue, but there is also a hint of curvature in the non-stellar

spectrum. The latter is consistent with the variability analysis in W12 that the amplitudes appears

larger in g and i-bands than in r band. But we cannot rule out the possibility that there is an

– 6 –

additional old stellar population in the SDSS spectrum. If the curvature is confirmed, the non-

stellar continuum may consist of two components, e.g., a disk component and a reprocessed outflow

component.

It was reported that J0952+2143 also has a non-stellar continuum in the SDSS spectrum (Ko-

mossa et al 2009). These authors fitted the SDSS spectrum with a single simple stellar population

plus a black-body emission. Our analysis does not confirm this. The SDSS continuum can be fitted

very well with a scaled up MMT spectrum, therefore, a non-stellar component is not needed. For

J1350+2916, W12 found that there is 3σ evidence for brightening in the g band between SDSS

photometric and spectroscopic observations. Our analysis does not require a non-stellar compo-

nent in this case. However, for this object, the MMT spectrum does not cover the 4000A break

or blueward of it, which is crucial to the detection of non-stellar component, because a non-stellar

continuum is most prominent in short band and also the 4000A break provides strong constraints

on the scale of the stellar component.

3.3. Broad Emission Line Variability

Three of the four objects displayed prominent broad Hα, Hβ or He IIλ4686 lines in their SDSS

spectra. Broad Balmer emission lines in J0952+2143 and J1350+2916 displayed double horns with

separations between 2000 – 3000 km s−1(see Komossa et al. 2008; W12). The FWHMs of these

lines are 2100 and 2600 km s−1, respectively. J0748+4712 showed two strong and broad (several

hundred A) bumps peaked around 4600 and 6560A. The broad bump at 4600A was interpreted

as blue-shifted He II by W11. With the more reliable starlight subtraction here, this line also

seems double peaked with the red peak close to the rest frame of He II while the blue peak at

−8, 000 km s−1, although the signal to noise ratio is still a bit low. It is hard to attribute the

blue peak to the contamination of another broad line as there is no strong lines expected in this

wavelength range. Although the double peaks are not evident in the broad Hα due to its weakness,

the data are consistent with this. Thus, broad lines in all three TDE candidates show double

peaked profiles. We have fitted the narrow Hα and [N II] double lines with three narrow Gaussians

and use a broad Gaussian to estimate the broad component of Hα line. Hα is detected with a flux

of 2.1×10−15erg cm−2 s−1. This gives a He II/Hα ratio of 3.3.

These broad bumps in J0748+4712 and double horns in J0952+2143 and J1350+2916 disap-

peared in the MMT spectra (Figure 3). Only very weak broad Hα and Hβ lines can be spotted

in the MMT spectrum of J0952+2143 (middle panel of Figure 3 and figure 4). Any broad lines, if

present, must be below the detection limits in the other two objects.

To measure the broad emission lines in J0952+2142, we fit the continuum subtracted spectra

using Gaussians (Figure 4). All narrow lines show symmetric profiles. The Hα and Hβ lines display

an additional redshifted broad component. Each symmetric narrow line is fitted with a Gaussian,

while one more Gaussian component is added for the redshifted broad component of each Balmer

– 7 –

line. Hα and Hβ are fitted, with the widths and centroids of each component locked. The line

flux are listed in table 2. The broad component is redshifted by about 120 km s−1 relative to the

systematic velocity and has an FWHM of 620±45 km s−1. It could be the relic of the fading broad

Balmer lines seen in the SDSS and NTT spectra that was described in Komossa et al. (2008 &

2009). In those two spectra, which were taken on December 30, 2005 and Feb 6, 2008, the broad

Balmer components had redshifts of 560 km s−1 and 270 km s−1, and FWHMs of 2100km s−1 and

1500km s−1. The line flux in the MMT spectrum is only 9% of that in SDSS spectrum and 25% in

NTT spectrum.

3.4. Coronal Line Variability

Figure 5 compares the coronal line spectra, after subtracting the continuum of MMT spectra

from those of the SDSS spectra over the wavelength range from 5200A-6450A, covering coronal

lines [Fe VII], [Fe X]λ6376 and [Fe XIV]λ5304. The one Gaussian fit result are also listed in table

2. A close look at the figure reveals some interesting variation patterns. Two (J1342+0530 and

J1350+2916) of three objects without [Fe VII] emission lines in the SDSS spectra now show these

lines in the MMT spectra, while in J0952+2142, [Fe VII] lines prominent in the SDSS and NTT

spectra become very weak now. But individual objects also show some different properties, that

will be further described below.

J0748+4712 – its SDSS spectrum shows strong high-ionization coronal lines ([Fe X], [Fe XI]

and [Fe XIV]) but no [Fe VII] lines. In the follow-up spectroscopic observation in 4-5 years after

the SDSS observation, all the coronal lines disappeared although the S/N ratios of the spectra were

very low (W11). The high S/N ratio MMT spectrum confirmed this, and there were no [Fe VII]

or higher ionization lines. This marks this object with the shortest duration of coronal lines.

J0952+2143 – its SDSS spectrum shows both high-ionization coronal lines and relative low-

ionization coronal lines ([Fe VII]). In the follow-up observations carried out with the Xinglong

2.16m telescope and the ESO NTT telescope 2-3 years later, all the coronal lines with ionization

higher than [Fe VII] disappeared and [Fe VII] lines show a marginal decrease (Komossa et al. 2008;

2009). In our MMT spectrum, only weak [Fe VII] lines are detected.

J1342+0530 — its SDSS spectrum shows strong [Fe X], [Fe XI] and [Fe XIV] but no [Fe VII].

In the MMT spectrum, all [Fe X], [Fe XI] and [Fe XIV] disappear but [Fe VII] lines appear.

J1350+2916 – The variability of coronal line spectrum is similar to J1342+0530.

In summary, for all these sources, the coronal lines are shifted from high to low ionization

species, or to a lack of coronal lines.

– 8 –

3.5. Other Narrow Emission Lines

W12 showed that conventional narrow-line ratios measured from the SDSS spectra place these

objects in the locus of star forming galaxies on the BPT diagrams (Baldwin, Phillips & Terlervich

1981; also Kewley et al. 2006). This suggests that these lines come mainly from star-forming

regions around the nuclei. Figure 8 demonstrates the line ratios measured from the MMT spectra

on the BPT diagrams. The line ratios are now very close to or above the line of demarcation

between extreme AGNs and the star-formation region. This is at least partially attributed to the

increase of [O III] from the nucleus as discussed below.

Changes in the BPT diagram must be interpreted with caution because the SDSS and MMT

spectra were obtained using different apertures. The MMT long-slit spectra are extracted from

a smaller aperture (1”×3”) than the SDSS fiber size (3” diameter circle), and thus contain less

light of the host galaxy. The emission lines from the extended star-formation region should be

weaker in MMT spectrum than in SDSS spectrum. Indeed, the low ionization lines, including [S

II], [N II], and narrow Hα, Hβ, are substantially weaker in the MMT spectra than in the SDSS

spectra(table 2), suggesting that they come from a large region. However, [O III] in the MMT

spectra is significantly stronger than in SDSS spectra (Figure 6 and table 2). The difference (360%,

33%,75% and 50% for J0748+4712, J0952+2143, J1342+0530 and J1350+2916) is larger than the

calibration uncertainty (typically 4% for SDSS and <5% for MMT spectrum). This suggests that

[O III]λ5007 had brightened since the SDSS observation. The increment must come from the very

nuclear region that is related to the tidal disruption event.

It is surprising that we measure an [O III] doublet ratio [O III]λ4959/[O III]λ5007≃ 1/2,

which is very different from the conventional 1/3, in J0748+4712 and J1342+0530 (Figure 6). To

check for potential contamination from other lines, we examined the NIST atomic database for

lines within 3A of [O III]λ49592. Only Ti Iλ4958.3, Fe IIλ4958.2, Pm Iλ4959.5 and Fe Iλ4961.2

are within 3A of [O III]λ4959. Fe II can be ruled out because it would produce much stronger

emission at 4889.6A from the same upper level, which is not detected. For a similar reason, Fe I

can be excluded because it predicts a stronger emission line at 4916.3A. Ti I and Pm I have the

same problem in addition to a very small element abundance. Therefore, contamination with other

lines can be ruled out. We also looked for potential telluric absorption lines and sky lines around

the redshifted line wavelengths and found none. Since [O III]λ4959 and [O III]λ5007 share the

same upper level, the different line ratios can be only explained by the radiation transfer effect,

which requires a substantial optical depth for the line. A detailed interpretation of this is given in

§4.2.

Narrow He IIλ4686 shows a more complicated variability pattern. In J0952+2143, strong

narrow He IIλ4686 seen in the SDSS spectrum becomes barely detectable in the MMT spectrum,

2Giving the similar profiles of [O III] doublet in these two objects, contamination by lines with larger offset relative

to [O III]λ4959 can be ruled out.

– 9 –

along with weakening of coronal lines and the broad lines. On the other hand, narrow He IIλ4686

remains unchanged or shows a small increase in J1342+0530 in the course of the decrease of high

ionization coronal lines and the appearance of [Fe VII]. The situation is not clear in the other two

objects because of either blending with the broad component in the SDSS spectrum or weakness

of the line in the MMT spectrum. Note that He IIλ4686 in both SDSS and MMT spectra are

systematically broader than other low-ionization narrow emission lines, such as Hα and [S II], which

mainly come from star forming regions (Figure 7). This is consistent with the expectation that He

IIλ4686 is powered mainly by the continuum flare, and star-formation makes little contribution to

it.

4. Discussion

4.1. Formation and Evolution of Broad Lines

In our sample, three targets (J0748+4712,J0952+2143,J1350+2916) show broad recombination

lines (Hα, Hβ or He IIλ4686) in the SDSS spectra and all these lines are double-peaked or strongly

blue shouldered. Such lines can be formed in biconical outflows launched by the super-Eddington

accretion flow, by a fast outflowing unbound stellar debris, or a ring of accreted debris, that are

photoionized by strong radiation from the accretion disk. Bogdanovic et al. (2004) showed that

photoionized debris produces only weak broad Hα line (L(Hα) ∼ 1037 erg s−1), which is much

smaller than the observed one, for the tidal disruption of a solar type star by a black hole of

106 M⊙, because the stellar debris is confined within only a small subtending solid angle to the

accretion disk. There are several possible way in which emission lines can be enhanced. First, if the

star is evolved, e.g., a subgiant, the debris may spread out more in the vertical direction because

an evolved star has a much large size, thus producing more reprocessed emission lines. Second,

if the star’s orbit mis-aligns with the spin direction of the black hole, the debris disk may receive

more light from the accretion disk. Finally, the line equivalent width is likely also dependent on

the mass of black hole. Small black holes may produce larger line equivalent widths because the

tidal disruption radius is smaller so the stellar debris has a larger covering factor. It remains to

be seen whether taking into account all these effects can reproduce the observed line equivalent

width. Thus we favor the biconical outflow model because it predicts a large line equivalent width

and double-peak profile which are more consistent with observed lines (Strubbe & Quataert 2009).

In either scheme, broad lines are expected to fade on the time scales of months to a year.

The fading is caused by both the declining of the ionizing continuum and the evolution of the

debris properties or outflows. The variation of the ionizing continuum is qualitatively understood,

a nearly constant luminosity at the super-Eddington phase and a power-law decay phase with an

index around −5/3 to −5/2 (Rees 1998; Laodato & King 2009). In either stage, the ionizing

continuum becomes softer. In a quasi-steady accretion system, it is expected that the outflow

weakens as the accretion rate decreases. Since the accretion rate varies on dynamic time scales, the

– 10 –

structure of the outflow is likely complicated. The disk initially launches strong outflows during the

super-Eddington phase, which is accelerated to a high velocity. As the accretion rate decreases, the

disk will produce a weak outflow, and the radiation acceleration is greatly reduced; after certain

time, it may eventually become smaller than the gravitational one, and then outflow decelerates in

the gravitational field.

In J0748+4712 and J1350+2916, we can only put upper limits on the lifetime of the broad

lines to 5 years. In J0952+2143, we witness the declining of the broad lines. In the NTT spectrum

taken 768 days after SDSS spectrum, the redshifted broad Balmer emission lines and double horn

profile are still prominent (Komossa et al. 2009). The line flux declined by a factor of only 2.8

in comparison with SDSS observation. This is slower than a ∝ t−4/3 or t−5/3 law if we take

the initial flare to have occurred near the SDSS photometric observation, 375 days before the

SDSS spectroscopic observation. In its MMT spectrum, which was taken 2187 days after the

SDSS spectrum, the double horn profile disappeared but the weak redshifted broad component still

exists. Taking into account the short recombination time scale of excited H atoms (≈ 105/ne yr,

Osterbrock & Ferland 2006), an ionizing continuum is needed to produce the broad Balmer lines.

The continuum may form through late stage accretion of residuary stellar debris in very eccentric

orbits, or by fallback of failed outflows. The broad line flux is a factor of 4 lower than in the NTT

spectrum. Noting that both the line width and the centroid offset decreases with time, this is

consistent with the decelerating outflow scenario.

Probably due to the relative small size of the accretion disk, outflows from both sides are

observed, leaving a double peaked profile, in contrast to the single-peak blueshifted profiles of CIV

in luminous quasars (Richards et al. 2011; Wang et al. 2011). The He II line in J0748+4712 is

not symmetrical around zero velocity at the source rest frame. This perhaps can be attributed to

the partially obscuration of the base of the receding outflow. The small separation between the

two peaks in the other two objects can be produced by either a large inclination to the collimated

outflows or a small outflow velocity. As noted in W12, J0748+4712 was seen by SDSS much earlier

than the other two objects when the disk emission is still strong and outflows show a high velocity.

With the decrease of the accretion power, the outflows are decelerated in the gravitational potential

of the black hole, leading to small outflow velocities in other two objects.

As discussed in W11 (see also Peterson & Ferland 1986; Gezari et al. 2012), the large He

II/Hα ratio requires an over-abundance of helium relative to hydrogen, which is interpreted by

W11 as the tidal disruption of an evolved star. Following Peterson & Ferland (1986) and assuming

a gas temperature of 2 × 104K, a minimum of nHe/nH = 0.75 is required by assuming that most

of helium is in He+2.

Most TDEs discovered by X-ray and UV flares do not show broad emission lines, while three

of four objects in this sample do. This may be attributed to selection effects. On the one hand, the

broad line objects are usually considered as AGN and rejected for further monitoring to reduce the

number of sources for repeated spectroscopic follow-up (Gezari et al 2009; van Velzen et al 2011;

– 11 –

Cenko et al 2012). On the other hand, coronal-line selected objects preferably lie in the gas-rich

disk galaxies, and they tend to host small black holes. The accretion rate is expected to be higher

than that of a high- mass black hole, and outflows should be denser and stronger. If broad lines

are formed in outflows, one naturally expect that TDE by small black holes produce strong broad

emission lines. More theoretical calculations are required to verify this.

4.2. Consequences of optically thick [O III] emission

4.2.1. Simple estimates

The [O III] line ratio, which is ∼ 2 rather than the expected 3, suggests that the lines are

optically thick. In this section we examine some consequences of this observation. This discussion

is preliminary because there is not yet sufficient spectroscopic constraints to fully define a model

for the observations.

If a line is thermalized, that is, the density is substantially above the critical density of the

line, then the level populations are given by the gas kinetic temperature and line photons are

lost in the scattering process because they are collisionally deexcited. The intensity of a line

is then given by Iν = Bν [1 − exp(−τ)] where Bν is the Planck function and τ the line optical

depth. The λλ5007, 4959 line optical depths τ are in a 3:1 ratio, so the intensity ratio is given by

I(λ5007)/I(λ4959) = [1 − exp(−τ)]/[1 − exp(−τ/3)] ≈ 2. This function is shown in Figure 9. A

line ratio of ≈ 2 corresponds to an optical depth τ(λ5007) ≈ 1.5.

The actual line ratio is a function of both optical depth and density. If the electron density is

well below the critical density of the 1D2 level that produces the [O III] lines then photons will be

remitted after absorption. They will simply scatter out of the cloud with no loss of intensity and the

3:1 line ratio is maintained. Line photons are only lost when there is a large probability of collisional

deexcitation, they are thermalized, following absorption. The lines will be well thermalized when

the density is ne ≫ ncrit(1D2) ≈ 6.8 × 105 t0.54 cm−3, where t4 = T/104 K. If the density is less

than this, a larger λ5007 optical depth would be needed to reach the same line ratio.

More O III lines would be needed to estimate the parameters for the [O III]-forming region.

A good detection of the λ4363 line would be important. These observations are not available so we

will simply assume that the λ5007 optical depth is of order unity, t4 = 104 K, and that the electron

density is of order the 1D2 critical density.

4.2.2. Column density and Compton depth

The line absorption coefficient is given by

α = 1.497 × 10−6fluλµm/uDop cm2 (1)

– 12 –

where flu is the absorption oscillator strength, λµm the wavelength in microns, and uDop the

Doppler width in velocity units. Assuming the transition probabilities given by Storey & Zeip-

pen(2000) we find α = 5.57 × 10−17u−1Dop cm2.

The Doppler width is given by the thermal width if turbulence is absent;

uth =√

2kT/m cm s−1 ≈ 12.8√

t4/mAMU km s−1 ≈ 3.2 t1/24 km s−1 (2)

where the expression is evaluated for oxygen, mAMU = 16. The absorption cross section becomes

α = 1.74 × 10−22 t−1/24 cm2 (3)

The column density required for τ(λ5007) > 1 is then

N(O2+) > 5.75 × 1021 t1/24 cm−2 (4)

The solar O/H ratio is 4.9 × 10−4 (Asplund et al.2009) so the total hydrogen column density at a

metallicity Z will be

N(H) > 1.2× 1025 t1/24 Z−1 cm−2 (5)

The corresponding electron scattering optical depth is then

τ(e) = σTN(H) > 7.8 t1/24 Z−1 (6)

where σT is the Thomson cross section. This has the important consequence that, unless the

metallicity is quite large (Z ≥ 10), the [O III] line-forming region is optically thick to electron

scattering.

This means that we may have underestimated the line width in Equation 2 since λ5007 photons

will scatter off the rapidly-moving electrons. The mass of an electron is 2.94× 104 smaller than an

oxygen atom so the line width is 171 times wider. The column densities are increased by√171 ≃ 13

if electron scattering dominates the line width.

The thermal width of [O III] λ5007 at 104 K would be ∼ 550 km s−1 in the electron-scattering

dominated limit. This is substantially larger than the observed widths, 200 and 255 km s−1 3,

showing that the lines are not broadened into the full electron scattering limit. At optical depth

τ(e) < 1, the scattered photons will produce an extended wing with a width dependent on the

optical depth and electron temperature, supposed on the primary emission profile (Laor 2006).

Thus the profile of high velocity wing can be used to constrain the properties of the scatter.

Unfortunately, from our data we can not confirm or rule out the existence of the broad wing of [O

III] emission lines. The lack of a prominent wing would indicate that the gas has a high metallicity

of order Z ∼ 10.

3This width is√

2σv.

– 13 –

4.2.3. Suggestions from photoionization models

The large column density suggested in Equation 5 places constraints on the radiation field

shape and intensity. In its simplest form the photoionization balance equation may be written as

(Osterbrock & Ferland(2006))

φ(H) = ne np αB dl = ne αB N(H) (7)

where αB is the hydrogen Case B recombination coefficient, φ(H) is the flux of hydrogen-ionizing

photons, and dl is the Stromgren thickness, the thickness of the H+ layer. This can be written

in terms of the dimensionless ionization parameter U , the ratio of densities of hydrogen-ionizing

photons to hydrogen,

U(H) ≡ φ(H)

n(H)c= 1.1

αB

cN(H) (8)

where we assume that helium is single ionized with solar metallicity, so that ne = 1.1n(H). Putting

in numerical values and assuming αB ≈ 2.6 × 10−13 t−0.84 , the limit in Equation 5 becomes

U(H) > 102 t−0.34 Z−1. (9)

This is a large ionization parameter. Photoionization models of strong-[O III] lined objects

are often fitted with U ∼ 10−2 − 10−1. For photoionization by a continuum that extends to high

energies, such as the SED of an AGN, the ionization of the gas is proportional to U (Osterbrock &

Ferland 2006) and little O2+ is present for such large values.

This is not the case if the SED is soft. If few ionizing photons are present at energies that

can ionize O2+ to O3+ then the required column density of O2+ will be produced. This occurs if

the SED is equivalent to a blackbody with T < 5 × 104 K, or if a more energetic SED is filtered

through in intervening absorbing column which removes high-energy photons. This brings in the

nature of the ionizing source, a fundamental question in these objects.

A very large U with a soft SED is consistent with the 104 K temperature we have assumed.

In photoionization equilibrium the gas heating is proportional to the photoionziation rate, which

is equal to the recombination rate. There is no direct dependence on the flux of photons or

the ionization parameter. A stronger flux of photons produces higher ionization but the same

photoionization rate, which is equal to the recombination rate, so the heating rate is constant.

4.3. Formation and Evolution of Narrow Emission Lines

Coronal lines and a significant fraction of [O III] must come from ambient gas photoionized

by the flare as they are variable on time scales of several years. Most previous TDE candidates

that were discovered via X-ray or UV flares do not have such features. The recent discovered TDE

candidate PS110jh with a strong broad He IIλ4686 line but weak Balmer lines as J0748+4712 do

– 14 –

not show coronal lines either (Gezari et al. 2012). Thus only a fraction of TDE shows coronal

lines, and the presence of coronal lines can be independent of the existence of broad lines. This

can be understood in the framework already discussed. The presence of coronal lines is related

to the gas distribution close to the supermassive black hole (W12), while the formation of broad

lines depends on the strength of outflows or the geometry of debris relatively to the disk radiation,

which is likely anisotropic. Although the nuclear gas environment and outflows are expected to be

correlated with the black hole mass in the local universe statistically, there are many outliers.

Komossa et al. (2008) postulated that coronal lines are echoes of the continuum flare by

distant gas. As discussed in W12, the variations of emission lines are rather complex in this scenario,

depending on the time evolution of the flare as well as the distribution of gas surrounding the massive

black hole. The emission lines at a given time come from different regions that are illuminated by

the continuum with different time advances. Thus, in the optically thin case, the optimal emission

region for a specific line is determined by competition between the time fading of the continuum,

the position dependent time lag, the r−2 dilution of incident continuum intensity, and the radial

dependence of density. Figure 10 shows the snapshot of ionization parameter distribution for a

model in which the ionizing continuum varies with time as a power-law declining phase L(t) ∝ t−5/3,

and a smooth density distribution n ∝ r−β. In the optically thick case, additional frequency

dependent attenuation has to be considered, which steepens the ionizing continuum at distant

region.

Even with the sparse sampling, the observations clearly require a different duration time scale

of coronal lines in different objects. Coronal lines in J1342+0530 lasted for a time scale of at least

ten years with a transition from high ionization coronal lines only to low ionization lines only,

while in J0748+4712, all coronal lines disappeared in less than 4-5 years. It is not clear what

causes the different time scale: the continuum light curve or the gas distribution. We do not know

what happened between the SDSS observation in 2004 and Xinglong 2.16m observation in 2009 for

J0748+4712. Was there a similar transition from high ionization coronal lines only to low ionization

coronal lines, or all coronal lines disappeared simultaneously? Even more, also for J1342+0530 and

J1350+2916, is there an intermediate state with both high and low ionization coronal lines as seen

J0952+2143? If so, then there is a continuous transition from high to low ionization state and we

can unify the high and low ionization coronal line emitter through time evolution. Unfortunately,

we do not have the data in these important evolution stages.

[O III] variability provides additional constraints on the gas distribution in the inner parsecs

and the evolution of the ionizing continuum. The brightening of [O III] follows the declining of

coronal lines, i.e., average ionization of line emitting gas decreases continuously. Giving the short

recombination times of O2+ (1.3× 105 (106 cm−3/ne) s), the gas is in quasi-ionization equilibrium.

This requires either attenuation of the ionizing continuum or/and the density must decrease more

slowly than r−2. The analysis in the last section suggests a soft ionizing continuum. Combining

with the strong coronal line emission in the early spectra, the supports the idea that hard ionizing

photons had already been filtered.

– 15 –

We have derived a lower limit on the gas column density in the last section for two sources

(J0748+4712 and J1342+0530) with non-canonical [O III] ratios. Bearing in mind that the flare

is less than ten years, so O2+ ions exist within the region bounded by the parabolic surface with

a lag of ten years (see Figure in W12). The large column density sets a lower limit on the gas

density of order n(H) ∼ 1.2 × 1025 t1/24 Z−1/c ∆t cm−3 = 2 × 106 t

1/24 Z−1 ∆t−1

6 cm−3, where

∆t6 = (∆t/6) yr.

Assuming that all observed [O III] comes from the same region, we can derive the volume and

mass of [O III] emitting gas. At a temperature 104 K, the volume emissivities of [O III]λ5007 are

2.0×10−16, 1.7×10−14, 6.7×10−13 and 9.5×10−12 erg s−1 cm−3 for n(H) = 104, 105, 106 and 107

cm−3 assuming a solar abundance. The observed [O III]λ5007 luminosities for J0748+4712 and

J1342+0530 are similar and around 9×1039 erg s−1, requiring a volume of line emitting gas of only

2.7×1052 cm3 for n(H) = 106 cm−3. This volume is only a small fraction of the volume swept by

the flare radiation in nearly 10 years, suggesting a small filling factor. Combining this with a large

column density implies that the emission region consists of only a small number of thick clouds.

4.4. Could the Powering Source be AGN Variability?

A substantial fraction of AGNs show coronal lines, but their strengths are usually much weaker

than [O III]. In the SDSS spectra, coronal lines of these TDE candidates are 1-2 orders of magnitude

stronger than that of strong coronal line emitting Seyfert galaxies. Further, large body of AGN

monitoring have shown that coronal lines in most AGNs are quite stable, while in rare cases which

coronal lines do vary on time scale of years and the amplification is usually moderate (Penston

et al.1984; Veilleux 1988). Only in one extreme case IC 3599, a strong decline of [Fe X] by two

orders of magnitudes was observed after an unusual soft X-ray burst by a factor of 100, that was

sometimes interpreted as tidal disruption event (Brandt et al. 1995; Grupe et al. 1995; Komossa &

Bade 1999). In our four TDE targets, coronal lines with ionization potentials higher than that of

[Fe VII] show strong fading of more than two orders of magnitudes within ten years and [Fe VII]

lines appear in two targets without them in SDSS spectra. That makes them very extreme if they

were AGNs. In addition, large amplitude variations (a factor of 100) in the highly ionized coronal

lines require large amplitude variations in soft X-rays. Up to now only narrow line Seyfert 1 galaxies

and BL Lac objects are known to show such large amplitude variability in X-rays (Eracleous et

al.2012). All the seven ECLEs targets are covered by FIRST survey (Becker et al. 1995), but only

J0938+1353 was detected with a low flux of 1.93 mJy, while others are below the detection limit,

typical 1 mJy. So from optical spectrum and radio flux, we can rule out both possibilities.

Along with the variation of coronal lines, a significant increase of [O III] emission lines is

detected in the four TDE targets. There is no clear evidence that emission lines such as [O III]

from normal AGN NLR would show strong variability due to its large size. So persistent AGN

activity is very unlikely the powering source of the variation of coronal lines and [O III] lines.

– 16 –

The disappearance of broad emission lines are also very different from other Seyfert galaxies

monitoring so far. In these Seyfert galaxies, broad emission lines usually vary in response to

changes in the continuum with a moderate amplitude. Only two cases were reported for temporary

disappearance of broad emission lines over a period of a few months accompanying with large

continuum variations in NGC 4151 and NGC 5548 (Penston & Perez, 1984; Iijima et al. 1992).

In J0748+4712 and J0952+2143, which have several observations after the outburst, we only find

broad lines are either fading or disappeared. Therefore, there is no evidence for re-starting of the

central engine. Considering all these facts, we believe that the powering source is likely a strong

outburst in an quiescent galaxy rather than stochastic variability of an AGN.

4.5. Could the Powering Source be Supernova Explosion?

For the Supernova (SN) scenario, it is unable to explain the high luminosities of coronal lines

and the absence of other low-ionization lines, and the continuum variability in some targets(W12).

Also, the MMT follow-up observations provide additional constraints on the SN scenario. First,

four TDE candidates show an increase of [O III] between SDSS to MMT observations, and the

[O III] are systematically broader than normal low-ionization lines such as Hα (figure 7). This

characteristic is different from other known SN. Very few young SNs display variable narrow [O

III] emission. Some SNs do show [O III] emission a few years to up 100 years after explosion, but

these lines are usually broad, accompanying low ionization lines of similar widths (Milisavljevic et

al. 2012). They are formed in the SN shell interacting with the stellar winds. Second, based on

the total energy in the high ionization coronal lines, W12 estimated a total energy in soft X-ray

of order 1050 ergs. The emergence of [Fe VII] lines in these two objects with only high ionization

coronal lines in previous SDSS spectra (J1342+0530 & J1350+2916) and the persistence of [Fe VII]

in J0952+2143 suggest that a strong UVX ionizing continuum is still seen by the line emitting gas.

The brightening of the He IIλ4686 line in J1342+0530 indicates a further increase of EUV ionizing

photons absorbed by line emission gas in this object, probably arising from large covering factor of

gas at the time lag. These variations of [Fe VII] and He IIλ4686 lines re-enforce the conclusion in

W12 that a very energetic flare is required to power the coronal line emission.

As shown in the last section, the presence of an optically thick [O III] emission region requires

large ionization parameters. In echo models, gas is photoionized by the continuum flare that took

place 5-9 years ago, so the size of emission line region is R[OIII] ≃ 1 − 3 pc. The peak luminosity

of ionizing continuum can then be estimated from the ionization parameter, the distance and the

density,

Lion = 4πR2[OIII]n(H)U(H)c〈hν〉 > 7.9× 1045t−0.3

4 Z−1Rpcn6(H) erg s−1 (10)

with a reasonable n6(H) = n(H)/106cm−3 ∼ 1, the luminosity would be 1045−46 erg s−1, which is

much more luminous than any supernova and is close to the Eddington limit for a 107−8 M⊙ black

hole. Finally, the life time of broad lines are at most several years, which is much shorter than

these in type II SN.

– 17 –

5. Conclusions

We have carried out follow-up MMT observations of seven extreme coronal line emitters that

were studied by W12. Three objects turn out to be persistently strong coronal line emitters. Among

them, two are likely star-forming and AGN composite or tidal disruption of a giant star, and one is a

narrow line Seyfert 1 galaxy. In the other four objects, all coronal lines with ionization higher than

[Fe VII] disappear. Thus, they are tidal disruption candidates. [Fe VII] lines faded in J0952+2143

and appear in the MMT spectra of two objects with higher ionization coronal line only in SDSS

spectra. [O III] doublets are brightened in all four objects. Thus we are still witnessing echoes of

the continuous decrease of the gas ionization in the emitting region. If this trend continues, the

object will appear in the AGN locus of the BPT diagram. Further monitoring of the emission lines

is needed.

We also detected non-canonical [O III] ratios in two objects. The column density of O2+ must

be large to make [O III]5007 optically thick. For reasonable metallicity, the H column density is

large and the gas is probably optically thick to electron scattering. Lines will be broadened due to

electron scattering, which can be tested with future high quality data. The ionization parameter has

to be very large to get this column density, and this also requires that the ionizing SED be relatively

soft, equivalent to a T < 5×104 K blackbody, or oxygen would be too ionized. A gas temperature of

order 104 K will occur for such extreme conditions due to the nature of photoionization equilibrium,

if the SED is this soft.

These observations can be fitted into a picture where giant molecular clouds are illuminated

by continuum flares with an ionizing continuum luminosity of order 1045−46 erg s−1 within a few

parsecs of the nucleus. The gas in the inner face of the cloud produces the coronal lines. As high

energy photons are absorbed by the gas, the ionizing continuum illuminating outer part of the

clouds softens and creates a thick [O III] emission region. The presence of giant molecular clouds

in the center parsec and no persistent nuclear activity is striking, suggesting that perturbation in

the gas is more fundamental for the black hole fueling. Clearly, we still miss some critical evolution

phases, which prevents us from probing a direct connection between these two subclasses of extreme

coronal line emitters. Future follow-ups of new discoveries of such objects from the spectroscopic

surveys of LAMOST and BigBOSS can fill the gaps and yield a more complete picture of emission

line evolution.

Broad emission lines previously seen in the SDSS spectra of J0748+4712 and J1342+0530

vanished in the MMT spectra, and there is only weak broad Balmer lines in J0952+2143. In

J0952+2143, we observed continuously fading of the broad lines by a factor of 11 in past 8 years as

well as a decrease of the line width and velocity shift. The line profile variability perhaps reflects

the deceleration of emission line gas in the gravitational field of the black hole due to a lack of

further radiative acceleration. With improved subtraction of stellar light, the broad lines in these

objects are found to all be double-peaked or blue-shouldered, suggesting it is a general property,

probably from biconical outflows.

– 18 –

We would like to thank Cai Zheng, Zhen-Ya Zheng and Hui Dong for their great help in

MMT observation and Stefanie Komossa for her very useful discussion. This work is supported

by NSFC 11233002 and 10973013. This research uses data obtained through the Telescope Access

Program (TAP), which is funded by the National Astronomical Observatories, Chinese Academy of

Sciences, and the Special Fund for Astronomy from the Ministry of Finance. Observations reported

here were obtained at the MMT Observatory, a joint facility of the University of Arizona and the

Smithsonian Institution. GJF acknowledges support by NSF (1108928; and 1109061), and STScI

(HST-AR-12125.01, GO-12560, and HST-GO-12309).

REFERENCES

Asplund, M., Grevesse, N., Sauval, A. J., & Scott, P. 2009, ARA&A, 47, 481

Ayal, S., Livio, M., & Piran, T. 2000, ApJ, 545, 772

Baldwin, J. A., Phillips, M. M., & Terlevich, R. 1981, PASP, 93, 5

Becker, R. H., White, R. L., & Helfand, D. J. 1995, ApJ, 450, 559

Bogdanovic, T., Eracleous, M., Mahadevan, S., Sigurdsson, S., & Laguna, P. 2004, ApJ, 610, 707

Brandt, W. N., Pounds, K. A., & Fink, H. 1995, MNRAS, 273, L47

Bruzual, G., & Charlot, S. 2003, MNRAS, 344, 1000

Cenko, S. B., Bloom, J. S., Kulkarni, S. R., et al. 2012, MNRAS, 420, 2684

Esquej, P., Saxton, R. D., Komossa, S., et al. 2008, A&A, 489, 543

Ferrarese, L., & Merritt, D. 2000, ApJ, 539, L9

Gebhardt, K., Bender, R., Bower, G., et al. 2000, ApJ, 539, L13

Gezari, S., Halpern, J. P., Komossa, S., Grupe, D., & Leighly, K. M. 2003, ApJ, 592, 42

Grupe, D., Beuermann, K., Mannheim, K., et al. 1995, A&A, 299, L5

Gezari, S., Basa, S., Martin, D. C., et al. 2008, ApJ, 676, 944

Gezari, S., Heckman, T., Cenko, S. B., et al. 2009, ApJ, 698, 1367

Gezari, S., Chornock, R., Rest, A., et al. 2012, Nature, 485, 217

Hills, J. G. 1975, Nature, 254, 295

Kewley, L. J., Groves, B., Kauffmann, G., & Heckman, T. 2006, MNRAS, 372, 961

Komossa, S., & Bade, N. 1999, A&A, 343, 775

Komossa, S., Zhou, H., Wang, T., et al. 2008, ApJ, 678, L13

Komossa, S., Zhou, H., Rau, A., et al. 2009, ApJ, 701, 105

Kormendy, J., & Richstone, D. 1995, ARA&A, 33, 581

Laor, A. 2006, ApJ, 643, 112

– 19 –

Iijima, T., Rafanelli, P., & Bianchini, A. 1992, A&A, 265, L25

Lodato, G., King, A. R., & Pringle, J. E. 2009, MNRAS, 392, 332

Lu, H., Zhou, H., Wang, J., et al. 2006, AJ, 131, 790

Magorrian, J., Tremaine, S., Richstone, D., et al. 1998, AJ, 115, 2285

Milosavljevic, M., Merritt, D., & Ho, L. C. 2006, ApJ, 652, 120

Milisavljevic, D., Fesen, R. A., Chevalier, R. A., et al. 2012, ApJ, 751, 25

Oliva, E. 1997, IAU Colloq. 159: Emission Lines in Active Galaxies: New Methods and Techniques,

113, 288

Osterbrock, D. E., & Ferland, G. J. 2006, Astrophysics of gaseous nebulae and active galactic

nuclei, 2nd. ed. by D.E. Osterbrock and G.J. Ferland. Sausalito, CA: University Science

Books, 2006

Penston, M. V., & Perez, E. 1984, MNRAS, 211, 33P

Penston, M. V., Fosbury, R. A. E., Boksenberg, A., Ward, M. J., & Wilson, A. S. 1984, MNRAS,

208, 347

Peterson, B. M., & Ferland, G. J. 1986, Nature, 324, 345

Rees, M. J. 1988, Nature, 333, 523

Richards, G. T., Kruczek, N. E., Gallagher, S. C., et al. 2011, AJ, 141, 167

Storey, P. J., & Zeippen, C. J. 2000, MNRAS, 312, 813

Strubbe, L. E., & Quataert, E. 2009, MNRAS, 400, 2070

Ulmer, A. 1999, ApJ, 514, 180

van Velzen, S., Farrar, G. R., Gezari, S., et al. 2011, ApJ, 741, 73

Veilleux, S. 1988, AJ, 95, 1695

Wang, H., Wang, T., Zhou, H., et al. 2011, ApJ, 738, 85

Wang, T.-G., Zhou, H.-Y., Wang, L.-F., Lu, H.-L., & Xu, D. 2011, ApJ, 740, 85

Wang, T.-G., Zhou, H.-Y., Komossa, S., et al. 2012, ApJ, 749, 115

This preprint was prepared with the AAS LATEX macros v5.2.

–20

–

Table 1. Basic Data of Extreme Coronal Line Emittersa

No Name z Mi,totb Obs Intervalc Broad Line [Fe VII] [O III] NSCd

SDSS MMT SDSS MMT

1 SDSS J074820.66+471214.6 0.0615 -19.75 2866 fading no no increase yes no

2 SDSS J095209.56+214313.3 0.0789 -20.41 2187 fading yes yes increase no no

3 SDSS J134244.42+053056.1 0.0366 -18.91 3548 absence no yes increase no no

4 SDSS J135001.49+291609.7 0.0777 -19.76 2073 fading no yes increase no no

5 SDSS J093801.64+135317.0 0.1006 -21.29 1834

6 SDSS J105526.43+563713.3 0.0743 -20.01 3548

7 SDSS J124134.26+442639.2 0.0419 -19.95 2859

aFirst four objects show significant variations in continuum and emission lines between two boservation, last three do not. We just list

the variations of the first four objects.

bMi,tot are estimated from SDSS photometry.

cDays bewteen SDSS and MMT observation.

dNSC stands for Non-stellar conitnuum.

–21

–

Table 2. Narrow Emission Line Fluxa of Variable Targets

No Hαn Hβn Hαb Hβb [Ne III] [He II] [N II] [O I] [O III] [O III] [O III] [Fe VII] [Fe VII] [S II] [S II]

λ3896 λ4686 λ6583 λ6300 λ4363 λ4959 λ5007 λ5722 λ6088 λ6716 λ6731

1 131(4) 49(4) < 2 < 6 48(4) 12(3) < 8 31(4) 65(4) < 3 < 4 23(4) 21(4)

2 108(7) 26(4) 155(10) 39(7) 31(4) 11(3) 64(5) 14(4) < 16 72(3) 215(3) 13(3) 17(5) 25(4) 27(4)

3 171(5) 27(4) 33(4) 49(4) 20(5) < 15 92(3) 182(3) 18(3) 31(4) 23(5) 17(5)

4 79(5) 15(3) 6(2) 22(4) < 5 < 9 22(1) 66(1) 8(2) 12(4) 9(2) 9(2)

aEmission line fluxes are in units of 10−17 erg cm−2 s−1. A superscript ’n’ means narrow line compoments and a superscript ’b’ means broad line

compoments.

– 22 –

Fig. 1.— MMT(black) and SDSS(red) spectra of seven extreme coronal line emitters from W12.

Continuum fit are plotted in green. Coronal lines and He IIλ4686 are marked in blue. The spectra

have been shifted in vertical direction for clarity.

– 23 –

Fig. 2.— SDSS spectra of J0748+4712, J0952+2143, J1350+2916 (black) and scaled up MMT

spectra’s 6ICs model (red) and the residuals of the substraction (blue). Both the spectra and

models are smoothed by 3 pixels.

– 24 –

Fig. 3.— Broad bump around 4600A in J0748+4712 (left panle,smoothed by 3 pixals) and Hα

regions of J0952+2143 (middle panel) and J1350+2916 (right panel). MMT spectra are plotted in

black and SDSS spectra are plotted in red, all the spectra have been continuum substracted.

– 25 –

Fig. 4.— J0952+2143’s Hα (low pannel) and Hβ (up pannel) line profiles(black) in MMT spectrum.

Normal narrow component fit is ploted in blue, redshift broad component is ploted in green. Model

sum is plotted in red.

– 26 –

Fig. 5.— Coronal line of four TDE targets. From top to bottom: J0748+4712, J0952+2143,

J1342+0530, J1350+2916. (black: MMT spectra, red: SDSS spectra)

– 27 –

Fig. 6.— [O III] lines of four TDE targets. Same order and color as figure 5.

– 28 –

Fig. 7.— Line width of Hα versus the width of [O III] (red) and He II (green) line in four TDE

targets in MMT spectra. Only lines with > 3σ detection are plotted. The straight line denotes

one-to-one relation.

Fig. 8.— Evolution of TDE targets in BPT diagrams (diamond: SDSS, square: MMT; blue:

J0748+4712, red: J0952+2143, green: J1342+0530, black: J1350+2916). Those targets evolve from

star-forming region towards extreme star-forming region and weak Seyfert region. The changes are

mainly due to the increasing of [O III] and smaller aperture of MMT spectra.

– 29 –9λlppe1]9a−l−1

t

tOl

i

iOl

c

λλλlppeλλ

tpc

pOpt pOt t tp

Fig. 9.— The dependence of the [O III] λ5007/λ4959 ratio on optical depth of λ5007. A line ratio

of ≈ 2 corresponds to an optical depth τ(λ5007) ≈ 1.5.

– 30 –

0.5 0.53 0.58 0.68 0.86 1.2 1.8 3 5.3 9.4 17 32

Fig. 10.— A snapshot of the ionization parameter around a black hole that contributes to the

observed emission lines at a specific time in an over-simplified model, in which the ionizing contin-

uum decreases with time as ∝ t−5/3 when t > t0 (where t0 is the time between the star was tidal

disrupted and most bound material returned to pericenter) and the density distribution decreases

as ∝ r−1(left panel) and ∝ r−2(right panel). The observer is to the left of the figure at time 10 t0.

The overlaid blue lines are intensity contours with an interval of two times fold.